> Bad performance for large OFFSET in SQL. When doing OFFSET Nin SQL, the database needs to scan and count N rows.
For user-facing apps, you’ve failed pretty hard already if users can’t find what they need on the first page or so.
If your offsets are big enough for this to matter, I’d rather spend time on why users are so many pages in to the data than optimizing the performance of later pages.
(Processing clients, on the other hand, should query for what they need. Processing in batches may make sens, so there cursor- or what they call keyset-based “pagination” makes good sense. Though in the case of processing clients, I wouldn’t call it “pagination”… it’s more like batches. I’ve mainly used “kelset-based” pagination for processing events, which can alleviate some of the “cons”.)
Users do fuzzy searches across data all the time. They may remember approximately when they last saw a record but not its exact name, or their recollection may be close but imperfect.
For user-facing apps, you’ve failed pretty hard already if users can’t find what they need on the first page or so.
If your offsets are big enough for this to matter, I’d rather spend time on why users are so many pages in to the data than optimizing the performance of later pages.
(Processing clients, on the other hand, should query for what they need. Processing in batches may make sens, so there cursor- or what they call keyset-based “pagination” makes good sense. Though in the case of processing clients, I wouldn’t call it “pagination”… it’s more like batches. I’ve mainly used “kelset-based” pagination for processing events, which can alleviate some of the “cons”.)